Title: MRF-based multi-view action recognition using sensor networks

Authors: Haitao Li

Addresses: Key Laboratory for Computer Virtual Technology and System Integration of Hebei Province, School of Information Science and Engineering, Yanshan University, Qinhuangdao 66004, China

Abstract: Action recognition has become an active area of research in the field of video surveillance. In this paper, a local space-time constraint Markov random fields (MRFs) model is proposed for the recognition of multi-view action based on the posture's articulation points from the sensor networks in the smart family space. The position distribution under different views and the time continuity of the posture sequence are used as the random field to label the corresponding action classes. Experimental results show that the proposed model can accurately recognise the actions of objects under multi-views in the family environment and requires low running time.

Keywords: action recognition; MRF; Markov random fields; multi-view actions; sensor networks; video surveillance; position distribution; posture sequences; posture features; visual perception; video images; simulation.

DOI: 10.1504/IJSNET.2017.083408

International Journal of Sensor Networks, 2017 Vol.23 No.3, pp.201 - 209

Received: 13 Sep 2016
Accepted: 13 Sep 2016

Published online: 27 Mar 2017 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article